3Delight Laboratory Thread: tips, questions, experiments

16061636566100

Comments

  • wowiewowie Posts: 2,029
    edited August 2016

    Could you give me some starting numbers? I tried your first example.

    Set Ubersurface base on a cylinder.

    Subsurface on, set to strength to 300% and backscatter boost to 10. Set opacity to 10%. Refraction on, 1.33

    Just to be clear, I'm using UberSurface2, not UberSurface. I think I've set the IOR to 1.1 on that render. My renderer options are pretty much default, except for progressive rendering, gamma correction (enabled) and ray trace depth set to 12. For an empty test scene, you can probably get away with 8.

    I'm still playing with the specular/reflection values, but those don't affect refraction and SSS. At least I think so.

    Here's another example, but this one with IOR 1.3 for the glass and the liquid. This is with US2 'MAT SSS Cream' preset, which loads the scatter/absorption values. The preset also changes Scatter strength to 7.38 and absorption to 0.010. I then set SSS strength to 300%, SSS scale to 10, backscatter to 10. This is with opacity set to 10%.

    As i noted before, if you want a murky refraction, you need to raise the opacity strength. I've changed the opacity strength to 20% and then lowered opacity color to 192,192,192.

    Here's another look where you can still see refraction through the glass and liquid.

    Higher opacity strength will make the SSS stronger (SSS is basically a diffuse response multiplied by the surface opacity). Lower opacity colors will make the refraction darker.

    I don't think SSS is the 'correct' way to mimic absorption be it in glass or water. Kettu's path length absorption is the more 'correct' way, but you still need extra parameters and a correctly modelled geometry. What I mean by parameters are depth information. I'm thinking the something like the Altitude and Depth Cue bricks can be used to 'guide' absorption when you want more absorption near the bottom/further inside and less near the top/nearer to the edge. However, that will still 'break' when you have a box of liquid that intersets with the container like my experiments with the pool in Stonemason's Pool House.

    SSS2.jpg
    826 x 1070 - 366K
    SSS3.jpg
    826 x 1070 - 359K
    SSS4.jpg
    826 x 1070 - 350K
    Post edited by wowie on
  • Oso3DOso3D Posts: 15,045

    I'd just like to get close, and having no luck.

    I tried your first example. My result looks like water. I... am very confused.

    (US2, cream, boosts that you mentioned, opacity 10%, refraction 1.33, path 12, etc)

     

    test.png
    1000 x 1000 - 1M
  • Oso3DOso3D Posts: 15,045

    Yeah, doing test after test, subsurface never makes a difference in the render. On looks same as off.

    wtf

     

  • wowiewowie Posts: 2,029
    edited August 2016

    Strange. Light intensity? I'm using an approximated physical light, so my intensity values are very high. However, even with the distant light at default settings you should be able to see something.

    Post your shader settings. Maybe there's a problem there. And what you're using for lights.

    US2 seems to install fine at least. If it's not, it'll render all black or white.

    Post edited by wowie on
  • Oso3DOso3D Posts: 15,045

    Diffuse color: Gray (128/128/128)

    Diffuse roughness: 1 (tried 0, makes no real difference)

    Indirect Diffuse Max Trace Depth: 1 (tried 12, makes no dif)

    Opacity Str: 20%

    Opacity color: White (tried various colors, no SSS effect, just colored volumes)

    IoR: 1.33

    Specular:

    Gray, 20% str, 85% gloss

    Subsurface color: White

    Sub Str: 300%

    Sub Refraction: 1.3

    Sub Scale: 1 (Set it to 10, 1, .1, at low values you get scaly look, but otherwise no effect)

    Backscatter boost: 10

    Subsurface group: 0 (tried 1, no dif)

    SSS Color: Blue (cream settings), Str 7.38

    Absorption: Yellow, Str .02

     

     

     

  • Oso3DOso3D Posts: 15,045

    You know, it occurs to me that part of the problem may be my angle of view and lighting. The camera is perpendicular to the light, when I think it's going to only be visible with the light in front or behind me.

     

  • Oso3DOso3D Posts: 15,045

    No, that still doesn't look right. Setting Sub Str really high, the water looks opaque. But I'm getting 0 color of any kind unless I set opacity color.

     

  • Oso3DOso3D Posts: 15,045
    edited August 2016

    Another example: Ambient white sphere with two spheres in it, each set to the Blood shader from (somewhere). It's an Ubersurface shader. AoA distant light (I've also tried standard light, same results)

    One side is normal, one side has Subsurface switched off.

    They look exactly the same.

     

    SSS comparison.png
    1000 x 618 - 382K
    Post edited by Oso3D on
  • wowiewowie Posts: 2,029

    Diffuse color: Gray (128/128/128)

    Diffuse roughness: 1 (tried 0, makes no real difference)

    Indirect Diffuse Max Trace Depth: 1 (tried 12, makes no dif)

    Opacity Str: 20%

    Opacity color: White (tried various colors, no SSS effect, just colored volumes)

    IoR: 1.33

    Specular:

    Gray, 20% str, 85% gloss

    Subsurface color: White

    Sub Str: 300%

    Sub Refraction: 1.3

    Sub Scale: 1 (Set it to 10, 1, .1, at low values you get scaly look, but otherwise no effect)

    Backscatter boost: 10

    Subsurface group: 0 (tried 1, no dif)

    SSS Color: Blue (cream settings), Str 7.38

    Absorption: Yellow, Str .02

    Can you post your scene file? Just create a scene with a primitive and your shader/lights setup.

  • Oso3DOso3D Posts: 15,045

    Here you go, and thank you very much for taking the time.

     

    duf
    duf
    3DL SSS experiments.duf
    152K
  • wowiewowie Posts: 2,029
    edited August 2016

    OK. I understand what's causing the confusion.

    SSS works just fine on that scene and on the cylinder. If you disable everything but the SSS, you can see it. Since it's quite weak, it gets 'buried' by the diffuse, refraction and reflection settings you used. Here's your scene with the default settings (i'm viewing the cylinder from the back so you can see the backscatter eaiser).

    Then i disable everything except SSS (diffuse, specular, opacity, reflection) so only SSS is enabled.

    If I disable SSS, it renders black - the way it should since there's nothing enabled in the shader.

    I also notice you set your SSS color to 255,0,0. That's a really big NO. Always set it to pure white (255,255,255). Use the scatter color and absorption color instead.

    Here's what I recommend as a starting point

    Should give you more visible SSS. Opacity color is pure white. Since that's the default, it's not visible under 'Currently Used'.

    Those settings will give you this:

    Without SSS, it'll look something like this

    Much more visible. Hope this points you in the right direction. One last note, I used a pretty high intensity on the light (1500%). If you're using much lower intensity (I saw you're using 200%), you need to boost the SSS even higher to compensate.

    test1.jpg
    826 x 1070 - 236K
    test2.jpg
    826 x 1070 - 218K
    test3.jpg
    826 x 1070 - 164K
    SSS.jpg
    826 x 1070 - 332K
    noSS5.jpg
    826 x 1070 - 337K
    SSSrefraction.jpg
    179 x 880 - 73K
    Post edited by wowie on
  • Takeo.KenseiTakeo.Kensei Posts: 1,303
    edited August 2016

    Handling of refraction shouldn't change from one renderer to the other

    For Iray use the caustic sampler

    I made two quick examples for Iray and 3delight. I have no difference in the refraction through the glass with custom shaders

    For the absorption, I don't think that it is needed for a liquid on a so short distance. Making unnecessary complicated things

     

     

    AA Iray glass 01.jpg
    611 x 892 - 284K
    AA delight glass 01.jpg
    611 x 892 - 287K
    Post edited by Takeo.Kensei on
  • wowiewowie Posts: 2,029

    Kettu,

    Here's a nice video about mesh and analytic lights in Renderman 21. The basically have the same light settings. wink

    The devs must've read my mind. laugh

  • Mustakettu85Mustakettu85 Posts: 2,933
    wowie said:

    What looks kind of odd to me is that the refractions of the background are all different between the renders. I'm assuming you use the same IOR value. If that is so, then all three renderers treat the value differently.

    Funny enough, the refraction of the grid looks close to the Luxrender shot.

    3Delight without the "cheat" and LuxRender do render the background refraction very similar. I don't have a DS4 LuxRender exporter, so I used DS3 and Tofusan's free one and naturally got the camera angle wrong. Refractions may well be identical between them when camera rotations are manually copied. But Iray just looks weird. 

  • Mustakettu85Mustakettu85 Posts: 2,933

    Handling of refraction shouldn't change from one renderer to the other

    I made two quick examples for Iray and 3delight. I have no difference in the refraction through the glass with custom shaders

    For the absorption, I don't think that it is needed for a liquid on a so short distance. Making unnecessary complicated things

    Hi! 

    You must have posted a different 3DL render than you intended - glossy refraction is just not cutting it as "no difference". Show us the real deal sans roughness, please.

    Supposing I want to animate the liquid, or just render a fluid sim mid-splash? There will be a difference between absoption and simply multiplying refraction by a colour. Especially if it's something like coffee.

    And since I get very little penalty for absorption - it uses the RT distance reported by that very trace() that does the GGX reflect/refract - I just don't see the advatange of complicating the user interface with even more switches. There is more use to RT subsets than only this cheat =)

  • Mustakettu85Mustakettu85 Posts: 2,933
    wowie said:

    Here's a nice video about mesh and analytic lights in Renderman 21. The basically have the same light settings. wink

    The devs must've read my mind. laugh

    Haven't yet had time to watch it all =)

    But they are at an advantage in that they write it all in C, including the renderer core, don't they =)

    I'll try to implement the rebalances that you suggested for delta lights, but what bugs me more is that I can't make PT and oldschool area lights give off a consistent intensity ratio when emitter area is rescaled. I can make, say, a 1 square meter plane emit the same energy in both modes, but when we get to scaling the plane, the intensity changes differently. To be honest, I'd just leave it as-is because using both types of area lights in a scene is a very unlikely scenario.

  • wowiewowie Posts: 2,029
    But they are at an advantage in that they write it all in C, including the renderer core, don't they =)

    Haven't gotten my hands on it yet. Non commercial version of RIS 21 isn't out yet. Renderer is probably C, but the shaders might be OSL.

    I'll try to implement the rebalances that you suggested for delta lights, but what bugs me more is that I can't make PT and oldschool area lights give off a consistent intensity ratio when emitter area is rescaled. I can make, say, a 1 square meter plane emit the same energy in both modes, but when we get to scaling the plane, the intensity changes differently. To be honest, I'd just leave it as-is because using both types of area lights in a scene is a very unlikely scenario.

    Well, my suggestion is to focus on the oldschool area light first. Just like that new integrator in RIS2, tag the new PT area lights  'experimental'. smiley

  • Mustakettu85Mustakettu85 Posts: 2,933

    But they're too good to be called that =) They're such an amazing workhorse. I really recommend using them over deltas and oldschool areas, especially for stills. Their only "downside" is that they don't work with traditional volumetrics like AoA's EasyVolume. Fairly easy to work around by using a spotlight set to a category that is _only_ "seen" by the volume and not other surfaces, anyway.

    "Oldschool" area lights I want to include more for backwards compatibility in case people want to use their material libraries based on traditional DS shaders. 

    Semi-OT: don't you wish DS would actually show you geometry normals? It's cool it can flip them, but outside of turning primitives into "interiors", this function is hard to use.

  • j cadej cade Posts: 2,310
     

    Semi-OT: don't you wish DS would actually show you geometry normals? It's cool it can flip them, but outside of turning primitives into "interiors", this function is hard to use.

    If you go under edit > preferences interface, you can turn backface lighting off which will make the backfaces black in the viewport. Its ugly for some stuff, but pretty useful for others.

  • Oso3DOso3D Posts: 15,045
    edited August 2016

    I'm pretty sure it's theoretical possible, but did anyone ever get around to making a shader that had elevation or curve elements to it, wrt terrain stuff?

    Basically, something Bryce/Carrara-like.

    There's at least one moss shader that might do it, although it's fairly specific. And then there's the projection shaders.

     

    Post edited by Oso3D on
  • wowiewowie Posts: 2,029
    Semi-OT: don't you wish DS would actually show you geometry normals? It's cool it can flip them, but outside of turning primitives into "interiors", this function is hard to use.

    Technically, DS and Poser never cared about normals :D You do that in a modeller app or a real professional grade, complete 3D app. In some twisted way, it's a good thing I think. When you have a problem, then it's generally traceable to bad modelling ie no thickness in glass, no beveled edges, not assigning the proper normals or using n-gons. You know, the usual stuff.

    Fairly easy to work around by using a spotlight set to a category that is _only_ "seen" by the volume and not other surfaces, anyway.

    Oh, did i just read that right? Kettu prefers a non physically correct workflow. smileyBut yes, generally speaking volumetrics needs a different approach. That's why there's OpenVDB and stuff. The older workflow and methods are really ugly hacks with extra helping of fudge. With those tools, volumetric effects actually look like the real thing.

    From Bertrand Benoit - http://bertrand-benoit.com/blog/1543-2/

    Here's another example I really like from Arion

    Noisy, but still awesome.

  • RAMWolffRAMWolff Posts: 10,249

    Hey Wowie,

    Have you ever thought of releasing a light set for sale here, completely set up for interiors and exteriors like many are doing with the iRay light sets?  You do amazing renders with 3Delight, every bit as good as any iRay renders I've seen but I can't ever get anything even remotely close to what you do and since I'm constantly making content, studying lighting isn't really high on my list.  If I could just clone myself ...... LOL 

  • wowiewowie Posts: 2,029
    edited August 2016
    RAMWolff said:

    Hey Wowie,

    Have you ever thought of releasing a light set for sale here, completely set up for interiors and exteriors like many are doing with the iRay light sets?  You do amazing renders with 3Delight, every bit as good as any iRay renders I've seen but I can't ever get anything even remotely close to what you do and since I'm constantly making content, studying lighting isn't really high on my list.  If I could just clone myself ...... LOL 

    Actually I have. laugh

    Most of my products come with a complete light set.

    The one I'm building right now is based of the one in Lumina Materials library, but recalibrated to get as close as i can to physical values.

    Reference - Maxime Roz with Indigo.

    UberSurface2, 3delight DS using the same HDRI.

    This is AO only, Not quite there yet, but getting pretty close.

    ref1.jpg
    1000 x 313 - 48K
    test1 (2).jpg
    998 x 998 - 480K
    Post edited by wowie on
  • RAMWolffRAMWolff Posts: 10,249

    Totally looking forward to this set then!  YAY! 

  • Mustakettu85Mustakettu85 Posts: 2,933

    I'm pretty sure it's theoretical possible, but did anyone ever get around to making a shader that had elevation or curve elements to it, wrt terrain stuff?

    Elevation, very much possible. I modified the basic "RiFog" example to use it, you need the shading point's Y coordinate and a function like smoothstep() to interpolate from one colour to another based on the coordinate. But my stuff is all RSL...

    There was an "altitude" brick in shader mixer. I think Pendraia did something with it, but I can't find her recipe. Probably it was in the old forums. You could ask her.

     

  • Mustakettu85Mustakettu85 Posts: 2,933
    j cade said:

    If you go under edit > preferences interface, you can turn backface lighting off which will make the backfaces black in the viewport. Its ugly for some stuff, but pretty useful for others.

    Thanks, I'll try this!


    wowie said:

    Technically, DS and Poser never cared about normals :D

    Tell me about this =) But it's so annoying to take models (and not necessarily old and/or free) into a modeler to fix those normals, isn't it, when technically DS can flip them.

    wowie said:

    From Bertrand Benoit - http://bertrand-benoit.com/blog/1543-2/

    Here's another example I really like from Arion

    Noisy, but still awesome.

    I guess that if BB has noise in a render, it's intentional. Kinda like those subtle chromatic aberrations.

    This reminds me that I keep meaning to try and hook 3DL OpenVDB shader to DS. Given that the only place the OpenVDB file enters into equation is the shader's parameter, it may well theoretically work (DS itself doesn't have to read it). What needs to be worked around is the colour array parameter because DS doesn't have a widget like this.

  • wowiewowie Posts: 2,029
    edited August 2016
    I guess that if BB has noise in a render, it's intentional. Kinda like those subtle chromatic aberrations.

    Nah, Bertrand Benoit didn't render the 2nd one. Probably the Arion devs or one of the users. Bertrand Benoit generally uses vray or Corona. He uses ArionFX for postwork though. The plugin just got updated yesterday.

    RAMWolff said:

    Totally looking forward to this set then!  YAY! 

    Hehehe. You'd have to bug Kettu to release her kit then. For that shot I used her 'old school' area light on some planes placed where the big lights are in the HDRI. The materials will still work with non area lights, but the highlight shape will be totally wrong if you're using plain point or spot lights. Unless you do some trickery with gels, but I haven't tried that method yet.

    Standalone have just been updated 12.0.113. 3dfM for Maya also gets an update, I"m hoping it works with Maya 2017.

    Btw, what is NSI? as in

    12.0.107 - 2016-07-11

    • OSL: Updated OSL libraries to the following head: d3abea49b89872bc1358714a
    • <strong>i-display</strong>: added graphical representation (docked view) of statistics inside the rendering window. Statistics are saved along with the file and can be viewed when the file is reloaded into i-display.
    • NSI: enable area light sampling for area lights declared using NSIMeshNode

    Seems like they've made lots of improvements to the OSL path tracer. smiley

    Post edited by wowie on
  • Oso3DOso3D Posts: 15,045

    Did anyone ever make a elevation/surface angle shader in 3DL?

    It'd make it a lot easier to do something 3Dl-ish with Terradome... hmm.

     

  • Did anyone ever make a elevation/surface angle shader in 3DL?

    It'd make it a lot easier to do something 3Dl-ish with Terradome... hmm.

    I think there were at least a couple of snow sahders that did this, Stonemason used one in Winter Kingdom as I recall and theer was also a Let it Snow product which I think was shader based. We also had some experimenting in the forums, though probably the oldest forum that is now lost.

  • Oso3DOso3D Posts: 15,045

    Oh right, the winter thing! I have that, I should try it.

     

Sign In or Register to comment.